Presenter: Md Tufajjal Hossain
Organization: New Jersey Institute of Technology
SAFETY
Abstract:
Heavy traffic volumes, frequent lane merges, toll plazas, and complex interchanges often create conditions for sudden and forceful vehicular stops, known as harsh braking (HB). Traditional safety studies rely on historical crash records, a reactive approach that delays countermeasures. Since HB events are continuously captured by connected-vehicle telematics, their spatial and temporal patterns offer a proactive surrogate for identifying crash-prone roadway segments. Therefore, this study evaluates the potential of harsh braking (HB) events as a surrogate measure of crash risk on New Jersey interstate highways. More than 8.5 million Drivewyze telemetry records and 45,000 police-reported crashes from July to December 2024 were analyzed. HB events were identified by a deceleration threshold of 6 ft/sec² (approximately 0.2g) and spatially matched to one-mile highway segments along with crash data. Descriptive analysis revealed strong spatial clustering of HB events and crashes along high traffic volume corridors such as I-95, I-80, I-78, and I-287, particularly near toll plazas and complex interchanges. Seasonal patterns showed HB counts peaking in late fall, coinciding with higher traffic congestion and adverse weather conditions. Statistical modeling using Negative Binomial (NB) and Zero-Inflated Negative Binomial (ZINB) regressions demonstrated a positive and significant relationship between HB events and crash counts. In the preferred ZINB model, the HB coefficient was 0.01 (p = 0.03), indicating that each additional HB event was associated with roughly a 1 % increase in expected crash frequency per segment. Although the per-event effect was modest, segments with repeated HB activity exhibited substantially elevated crash risk; for instance, an increase of 10 HB events correspond to an expected crash frequency of about 10 % higher. These findings demonstrate that crowdsourced telematics can serve as a practical, proactive tool for highway safety management, supporting early detection of high-risk locations and guiding countermeasures such as improved signage, targeted enforcement, and geometric enhancements before crash records accumulate.
Md. Tufajjal Hossain is a Ph.D. student in Transportation Engineering at the New Jersey Institute of Technology (NJIT). His research focuses on traffic flow modeling, intelligent transportation systems, and AI-driven traffic safety analysis. His recent work includes developing real-time incident detection models using crowdsourced Waze data and designing a data-driven framework for optimal Safety Service Patrol route identification based on historical crash data. He also explores crash severity prediction using large language models to enhance roadway safety analytics. At NJIT, he serves as a Teaching Assistant and has contributed to NJDOT-funded research at the Intelligent Transportation Systems Research Center. He is the recipient of the 2025 ITSNJ Outstanding Graduate Student Award and the Best Poster Award at the 2024 ITSNJ Annual Meeting, recognizing his academic excellence and contributions to advancing intelligent and data-driven transportation systems.
Presentation Slides:
